A dashboard does not have to look broken to mislead a business. It can load quickly, show plausible trends and carry the authority of a familiar management report while its figures conflict with those used by finance, operations or commercial teams.

An obvious error prompts investigation. A credible number can move straight into a forecast, budget or investment decision before anyone asks how it was produced.

The right question is therefore not simply whether the dashboard works. It is whether the number can be traced, explained and reconciled. That means examining the business definition, the data behind it and the process that turns that data into a reported measure.

Why trusted dashboards become unreliable over time

Reporting reliability rarely collapses because of one dramatic failure. It usually erodes as reasonable short-term decisions accumulate.

A spreadsheet fills a gap in a source system. A monthly export needs manual cleaning before it can be loaded. A recurring adjustment remains in one person's working file because it once helped the team meet a deadline. Each decision may be understandable in isolation, but together they create a reporting process that depends on undocumented knowledge and manual intervention.

Technology improvements can add another layer of uncertainty when old reporting paths remain in use beside their replacements. A new source system may provide better operational data, while downstream dashboards still apply earlier definitions or transformation rules. The platform has moved forward, but reporting has not fully caught up.

As data volumes grow, manual checks become less effective. People change roles, workarounds lose their original explanation and the reporting layer becomes more sophisticated. The result can be a dashboard that looks increasingly polished while the evidence beneath it becomes harder to defend.

How to investigate conflicting dashboard figures

Ardent starts by establishing what each number is intended to measure. The label on a metric can be as revealing as its calculation.

In subscription reporting, for example, a customer may cancel a renewal but retain access until the end of a paid period. One report can classify that customer as cancelled while another continues to count the customer as active. Neither figure is necessarily the result of a processing fault; the reports may be answering different questions.

Trials, payment grace periods, renewals and post-cancellation access create similar distinctions. These are business decisions that must be agreed before technical reconciliation becomes useful.

A recent Ardent data-platform assessment found that conflicting subscription reports arose primarily from competing definitions rather than a fundamental failure in ingestion or transformation. The immediate priority was not to force every total to match. It was to define what each measure should represent, then test whether the data and reporting logic applied that definition consistently.

Processing faults still need to be ruled out. A duplicate record in customer or product reference data can cause a transaction to be counted more than once when datasets are joined. Top-line totals may still appear plausible, particularly when errors offset each other, while figures by region, product or customer group are wrong.

What the business sees — Possible cause — What Ardent investigates

Finance and operations report different totals — Different timing rules or metric definitions — The agreed business definition, reporting period and treatment of exceptions

The overall total looks plausible but regional figures differ — Duplicate reference records or leaking data — Record matching, transformation logic and aggregation at each reporting level

A historic trend appears continuous — The underlying level of detail changed over time — Whether older aggregate data can be compared fairly with newer customer or transaction records

The dashboard refreshed successfully but the figures are stale — An upstream source arrived late, or a pipeline completed with incomplete inputs — Source freshness, completeness checks, pipeline dependencies and reconciliation to expected totals

This creates a structured basis for comparing source records, transformed warehouse data and dashboard totals. It shows where a difference enters the process, whether it represents a valid business treatment or a data fault, and who has the authority to resolve it.

When reporting metrics no longer reflect the business

Some of the most important reporting problems contain no calculation error. The pipeline processes the data correctly and the dashboard displays exactly what it was designed to show. The problem is that the organisation has changed while the measure has stayed the same.

Consider a business moving from one-off sales towards subscriptions or longer-term customer relationships. A dashboard centred on new customer acquisition may continue to show growth while saying little about renewals, churn or the value retained from existing customers. The acquisition metric remains valid, but it no longer gives leadership a complete view of performance.

Trusted reporting is therefore both a systems issue and a people issue. Ardent traces how data is collected, transformed and reported, and speaks to the people who use it. Product changes, new commercial models and shifts in customer behaviour can all change what the business needs to understand.

Technical teams should not be left to infer those changes from requests to amend a dashboard. Business leaders need to agree which measures now matter, what they mean and how they should be used.

Why historical trend data may not be comparable

Ten years of trend data does not necessarily mean ten years of comparable evidence.

A continuous chart can conceal changes in the data underneath it. Older reporting may contain only monthly or platform-level totals, while newer systems hold individual customer, subscription or transaction records. The line on the chart remains unbroken, but the level of detail supporting it has changed.

That matters when redefining a metric or comparing customer behaviour over time. Aggregate history may support a reliable view of total revenue, for example, without supporting a detailed analysis of retention by customer segment. Reporting should make that boundary clear rather than presenting unequal evidence as if it were directly comparable.

Data ownership and governance for trusted reporting

Reliable reporting requires clear ownership of business definitions, data processing and dashboard delivery. These responsibilities are connected, but they are not the same.

The person who can restart a failed data job is not necessarily authorised to decide whether a customer counts as active. Separating those responsibilities gives each issue a clear owner and prevents technical teams from quietly inheriting unresolved commercial or financial decisions.

Ardent's data assessments also examine whether platform knowledge and delivery responsibility are concentrated in too few people. That creates continuity risk when someone is unavailable and limits the team's ability to improve the platform while maintaining day-to-day reporting.

The consequences can extend beyond a dashboard. The same customer data may feed segmentation, marketing attribution, customer communications and access decisions. An inconsistent definition can affect both the figure leadership sees and the way the organisation treats a customer. A governed data model provides a consistent foundation for these uses.

Controls also need to keep important figures open to challenge. In December 2024, Macy's disclosed that an employee had concealed approximately $151 million in delivery expenses. Its subsequent annual filing identified weaknesses involving manual entries and reconciliation. The circumstances involved deliberate misconduct and differ from the operational issues Ardent typically encounters, but they show why material figures need to remain traceable, reconcilable and subject to review.

Define the data strategy before replacing technology

One of the valuable outcomes of an Ardent data assessment is identifying what does not need replacing. A core data platform can remain viable while reporting confidence is weakened by inconsistent definitions, limited controls or poorly understood connections between existing components.

In these cases, a wholesale rebuild would add cost and disruption without resolving the underlying disagreement. Ardent focuses investment on the changes the evidence supports, preserving useful infrastructure where appropriate.

Architecture simplification still requires judgement. Two integration tools may appear to perform similar jobs, but one may support historical reports, a specialist source or an important downstream process. Removing it without mapping those dependencies can interrupt reporting or create new gaps.

The data strategy should therefore establish agreed definitions, reconciliation and quality controls before deciding what to retain, consolidate or retire. Technology changes then follow a clear view of the problem rather than leading it.

Operational controls that keep dashboard data reliable

Resolving a disputed definition does not protect tomorrow's report from late, missing or duplicated data. Reporting also needs operational controls that show whether the underlying information is complete and current.

A successful pipeline run or recent dashboard timestamp provides only part of that assurance. Effective monitoring should check:

• whether expected source data arrived on time

• whether critical fields and identifiers are complete

• whether duplicate or unmatched records exceed agreed limits

• whether transformed totals reconcile with source systems

• whether changes to important reporting models passed controlled tests

Recovery must also have clear boundaries. A temporary connection failure may justify an automated retry. A late file can trigger processing once it arrives and passes validation. An unexplained adjustment, changed source structure or disputed business definition needs investigation rather than an automated change that merely makes two totals agree.

Good recovery identifies what failed, which data was affected and whether the resulting information is safe to distribute.

What trusted business reporting looks like

The strongest reporting environments are not those where nobody questions the numbers. They are the ones where important questions can be answered without reconstructing years of undocumented decisions.

Finance should be able to explain why its treatment differs from an operational view. Data teams should be able to trace a material figure from the dashboard back through transformation logic to the source. Business owners should know who approves a definition and when it was last reviewed.

That is the standard Ardent works towards with clients: reporting that shows what the number is, how it was produced and why it deserves to be trusted.

The most dangerous dashboard is the one whose polish discourages a second question. When a figure shapes forecasts, investment, operational priorities or customer commitments, looking right is not enough. It must stand up to scrutiny.

How Ardent can help

Ardent helps organisations trace the figures behind their reporting, resolve conflicting definitions and strengthen the controls that keep data reliable as the business changes. Our assessments cover the technology, the data and the way people define, own and use important measures.

If different teams are working from different versions of the truth, or confidence in reporting depends on a small number of people, speak to Ardent about a data-platform and reporting assessment.

Frequently asked questions

Why do two dashboards show different numbers

Two dashboards may use different source systems, reporting periods, filters or business definitions. The difference is not always a technical error. A structured investigation compares the definition, source data, transformation logic and final output to identify where and why the figures diverge.

How can you tell whether a reporting issue is caused by bad data or different definitions

Start by agreeing what the metric is intended to mean. Then reconcile source records, transformed data and dashboard totals using that definition. This separates valid differences in timing or treatment from duplication, missing data, incorrect joins or transformation faults.

What is data reconciliation

Data reconciliation compares figures at different stages of the reporting process to confirm that data has moved and changed as expected. It commonly checks source-system totals against warehouse models and dashboard outputs, while documenting any valid differences.

Should we replace our BI platform if users do not trust the dashboards

Not necessarily. The BI tool may be working correctly while the underlying problem sits in metric definitions, source data, transformation logic, controls or ownership. An assessment should identify the cause before the organisation commits to replacement technology.

How can organisations improve confidence in business reporting

Confidence improves when important metrics have agreed definitions, clear owners and documented sources; data is reconciled through the reporting process; pipelines are monitored for freshness and quality; and changes to critical models are tested before release.

Related Ardent insights

Conversational Dashboards: Interacting with Enterprise Data Using AIWhy meaningful questions depend on trusted definitions, governed data and reliable reporting foundations.

Intelligent ETL: From Data Preparation to Decision IntelligenceHow ETL is evolving beyond data preparation to support reliable business decisions.

Why Data Pipelines Must Become Intelligent SystemsHow data pipelines can reconcile complex sources and maintain dependable outputs.

References

Macy's third-quarter 2024 results and investigation findingsandMacy's 2024 annual filingon internal control weaknesses.